◐ rebuilding Bitcoin chain analytics are being rebuilt after a transaction-parser fix. Historical figures shown may be inaccurate until the re-ingest completes.
⬡ SUBSTRATE 🔒 sovereign Contact ▸ ⌘K research.semurg.io
warming the live Bitcoin value-flow graph, first transfers landing…
live Bitcoin value-flow · 0 addresses · 0 recent transfers · watch the money move

One 64-byte-container substrate does all of these

The whole industry assumes you need a separate engine per workload: a graph database, a search engine, a vector store, a cache, a warehouse, a tick store, a document store. This one substrate does every one of them, measured live on this box, on public data you can re-pull over Bitcoin P2P and re-run. Each row shows the measured number and a link to the demo that runs it for real. Pick any and prove the number yourself.

Live Bitcoin value-flow graph · whole-store size · warm traversal rate
1.4 s380.0M edges/s warm model
The live value-flow graph holds 1,721,098 transactions and addresses and 7,371,884,554 value-flow edges, the whole store across all shards. The node figure is the length of the shared content-addressed identity store, so every unique transaction and address is counted exactly once. The edge figure is the running total the ingest accumulates as each batch of blocks appends across all shards, started from whichever is higher of the durable checkpoint and one whole-store cross-shard fold taken at start-up. The figure above it is the measured WARM k-hop traversal rate over the resident working set, on this box. · the graph exceeds this box's RAM, so a traversal is disk-fed, not an in-memory cache. A single-call whole-store genesis-to-tip sweep still re-folds the co-located image on every call, which is far too slow to run on a page load, so this headline is the warm working-set rate and not a whole-store rate. The co-located image image that was meant to remove that re-fold has been deleted; co-location is the path being taken instead.
graph traversal · the canonical figure
468.3M TEPS warm
warm · in-process · single-shard · all-cores measured
The Graph traversal row below reads this same live figure and tier. The reproducibility API and the graph demo carry the same method label, so the number never disagrees with itself. see the proof ▸
Category Dataset Demo Semurg What it is
Graph traversal Bitcoin value-flow graph ▸ 468.3M TEPS● LIVE genesis-to-tip on the live 685M-edge chain, no ETL to load it
Recursive Bitcoin value-flow sqlrecursive ▸ 150.0M TEPS● LIVE recursive-CTE / multi-hop fund-trace
Graph algos Bitcoin value-flow gtriangles ▸ 90.0M wedges/s SIMD wedge + PageRank, on the live chain
OLTP point read Bitcoin tx / address sqloltp ▸ 623.8K SELECT/s● LIVE point read, durable on disk
SQL joins Bitcoin value-flow sqljoins ▸ 538.0M rows/s● LIVE HTAP join, on one copy, no ETL
KV / memory Bitcoin address keyspace kv ▸ 556.1K random GET/s● LIVE random point-GET, durable on disk, no network hop
Wide-column Bitcoin value-flow widecol ▸ 31.4M reads/s● LIVE wide-column reads
Document Bitcoin transactions document ▸ 33.0M joins/s● LIVE document join
Search Bitcoin tx / address search ▸ 7.45 GB/s● LIVE ad-hoc full scan, no index to build
OLAP / warehouse Bitcoin value-flow olap ▸ 314.0M rows/s● LIVE HTAP scan, on one copy
Time-series Binance live ticks tsdb ▸ 42.0M samples/s● LIVE range scan + bucket, on one copy
Vector Bitcoin address vectors vector ▸ 100% recall● LIVE recall = 1.0 at 4x density, same copy
Geo Bitcoin value-flow geo ▸ 314.0M points/s no separate spatial index
Ledger Bitcoin append log ledger ▸ 2.9M entries/s append-only, disk = truth
Feature store Bitcoin address features featurestore ▸ 30.5M reads/s online + offline features, one copy
Streaming Binance live stream streaming ▸ 2.5M events/s ingest and serve on one copy
Log / SIEM Bitcoin tx log logsiem ▸ 3.0M lines/s index-free grep-the-corpus
Object Bitcoin transactions object ▸ 2.59 GB/s bytes + graph + search in one engine
Every number is measured live on this box, on public Bitcoin and Binance data you can also download and re-run. No hand-picked comparisons, no invented multipliers. The numbers are the point.
every demo reports the full metric set cold · warm · hot
the all-metrics block
Each live demo renders one shared panel: writes, cold / warm / hot reads, traversals and scan throughput, plus cold / warm / hot hit-rate and RAM utilization (the resident working-set against the full dataset on disk, the conveyor-belt proof that RAM is a pipe, not the store). This proof is only fully honest on data larger than RAM, so a cold read cannot be faked by caching: the whole Bitcoin chain (over 123 GB, larger than this box's memory) is being ingested for exactly that, where the cold full-table scan physically cannot hide in RAM and you watch the conveyor stream it off NVMe at bandwidth. A tier reading measuring… means that tier's telemetry is still warming or its NIF has not reported yet, not a failure, run the demo again and it fills in. Open any demo and run it to see it.